v0.1.5
Proper markdown parsing, quiet mode, tech term pronunciation
Markdown
Replaced hand-rolled stripper with pulldown-cmark. Links now keep text and drop URLs, code blocks and images are skipped, HTML is stripped.
Quiet mode
-q / --quiet suppresses phoneme chunks and progress messages. Errors still print. Great for piping docs through voice:
curl -s https://automerge.org/llms-full.txt | head -20 | voice -q --markdown
Tech term pronunciation
Built-in substitutions for commonly mispronounced terms: JSON ("jay-sahn"), YAML, TOML, WASM, OAuth, NGINX, PostgreSQL, CRDT, SQLite, iOS, macOS. Always active, overridable with --sub.
Docs
- README: LLM-friendly design section, substitution docs, builtin voices table, architecture notes
- SKILL.md: lightweight reference card for AI agents
- voice-cli crates.io page now shows the root README
Previous
v0.1.4 — parallel startup, streaming playback